1. The Issue, Briefly

An attractive view about the referential mechanism of “now” is the following:

In this article, I argue against the Simple Token Reflexive view by showing that it relies on false assumptions about the metaphysics of certain speech acts. I then introduce the notion of a detonation condition, and propose the following metasemantics: when such a condition is met, the token detonates, and it picks out whatever its character makes it pick out. This remains a token reflexive view, but one that by-passes the metaphysical issues that plague the Simple View.

This Simple View is attractive for its simplicity, since it makes it quite easy to explain how speakers learn to use indexicals like “now”. It also offers an easy answer to the Humpty Dumpty problem for context-sensitive terms. The problem is this: if we allow the speaker’s intentions to determine the referent of these terms with no limits on what they can intend, we seem to be allowing the speaker to refer to whatever they want with their terms.Footnote ² Humpty Dumptyism is widely thought to be implausible, for instance because it makes successful communication hard, if not impossible, to explain: how could the hearer possibly figure out the speaker’s intentions, if they are unconstrained? Since the Simple Token Reflexive view does not take into account the speaker’s intentions, there is no threat of Humpty Dumpty (unless they determine when the speech act takes place, a worry that we will return to later).

This Simple Token Reflexive view works best if it is obvious when (and where) speech acts take place, and this is at least as clear as our intuitions about the referents of tokens of “now”. However, there is a class of speech acts for which these conditions fail, typically called “answering machine cases”. Consider this case: I write a note for my students, which I stick to my office door. It says “I’m not in right now”. I take it that every time that the note is read, “now” refers to the time of the reading, at least while I choose to leave the note on the door. While I am clear on this intuition, I am less clear on the following questions: when did I say that I was not in? When I wrote the note? Every time that the note is read? All the time while the note is up? I just don’t have an easy answer to these questions, and, as far as I know, there is no theory that offers widely accepted answers either. Indeed, Kaplan (Reference Kaplan, Almog, Perry and Wettstein1989b) and Predelli (Reference Predelli2005) have taken this as a reason to keep the semantics of natural languages away from talk of speech acts and tokens. I will set these worries aside for the purposes of this article.

Here is the problem for the simple token reflexive view: our intuitions about the locations of speech acts are unclear in answering machine cases, and in these cases, we are unable to explain how tokens of “now” get their referents. In addition, the metaphysics of speech acts will not be helpful either, as I argue in §3. However, this does not show that we should give up token reflexive views about “now” altogether.

The aim of this article is to develop and argue for a different token reflexive view about answering machine cases, one that is well placed to explain the behavior of context dependent expressions because it grounds a semantical theory in a wider discussion of speech acts, and of actions in general.

2. The View

Let me sharpen the problem a bit, and then formulate the view I will defend.

The class of cases I address is fairly large. To focus on their inner workings, let me describe a slightly simpler speech act than the one above: at t₁, A writes a note to B. The note says: “It’s raining now, so I’m not biking to work”. At a later t₂, B reads the note, and, knowing that A must have left it for B before leaving for work at t₁, is thereby informed that it was raining at t₁, and that A would consequently not bike to work. I would say that A said that (and told B that) it was raining at t₁. Now ask yourself these questions: when did A say that it was raining? And when did A tell B that it was raining? I do not have clear intuitions here, and for any person who has reported clear intuitions, there are others who have clear opposing intuitions. I have an even less clear intuition about the corresponding “where” questions. But I do have a clear intuition that A’s use of “now” refers to t₁.

Suppose that the reader grants my last two claims: that we lack clear intuitions about the where and when questions about the speech act, but that we do have clear intuitions about the referent of “now” as it occurs in the sentence used for the speech act. This takes some of the wind from the sails of the simple token reflexive view, since it purports to explain the latter by appealing to the former, and yet we are on intuitively firmer ground with the latter than the former.

This is not a knock-down argument, of course. Even if our intuitions are less firm for the explanans than for the explanandum, independent arguments could overcome that initial problem. But what else have we to go on here than our intuitions? The metaphysics of speech acts will have to take our intuitions, and the lack thereof, seriously too. Furthermore, we will see that there is an independent literature in the theory of action that is very relevant here, and we find just the lack of consensus we might expect given the unclarity of our intuitions.Footnote ³

One might retreat to the claim that these uses are not within the scope of semantics (Recanati (Reference Recanati and Burton-Roberts2005), Voltolini (Reference Voltolini2006), Stevens (Reference Stevens2009)). I will not argue directly against this option. For me, this option becomes attractive only if all else of a semantic nature fails. Since I do have a positive, token-referential story to tell, I take this to undermine an important rationale for putting these uses outside of semantics.

If tokens of “now” do not get their referent from the time of the speech act, where do they get it from? The solution will rely on the notion of detonation conditions. Here is the view I will defend, as applied to “now”:

Before we get to the details, I need to set the stage. The cases I consider are commonly called “answering machine cases” (Kaplan Reference Kaplan, Almog, Perry and Wettstein1989b). They are examples of language uses that are not face-to-face, instantaneous, or located in just one place. Typical examples are notes left for someone to read later, phone calls over large distances, and, as the name suggests, answering machine messages, whereby the owner of a phone line may communicate that they are not at home by the machine’s playing a message saying “I am not here right now”. I am using these cases specifically to probe token reflexive views, and to point out similarities to non-linguistic cases in action theory, to which we now turn.

3. Answering Machines and Killings

The fact that we seem unsure about the when and where in answering machine cases is not as unusual as it might seem. In this section, I point out that this is a feature of all action that takes some time, and whose realization depends on certain things happening later in that time frame.

As a classic example, consider a debate about the location of a very different type of action: killing.Footnote ⁴ Many conflicting answers have been proposed, and philosophers usually acknowledge some unease about the fit between their views and their pre-theoretical intuitions.

Consider this killing: C shoots D at time t₁ and place p₁, while D is standing at a different place, p₀. D dies from the wound at t₂ and p₂, where all the times and places are different from each other. Now ask yourself: when and where did C kill D?

According to Davidson (1969/Reference Davidson2001), C killed D at t₁ and p₁. He claimed that a killing is just an event that causes a death. This view has the counterintuitive consequence that D was killed (at t₁) before dying (at t₂).

To avoid this consequence, Thomson (Reference Thomson1971) argued that C killed D in the whole t₁ – t₂ interval. But this makes it the case that, although C was not doing anything related to D after t₁, he still counted as killing D all that time, until t₂.

In response, Bennett (Reference Bennett1973) proposed a refinement of Davidson’s view: C killed D at t₁ and p₁, but that only became a killing after t₂. This does indeed get rid of one issue with Davidson’s view, but leaves another, namely that D would still count as having been killed before dying, now that D is dead.

In 1998, Pietroski (Reference Pietroski1998), developing a view proposed earlier in Hornsby (Reference Hornsby1980), argued that we should distinguish between the action of killing, which is just the action of trying to kill, and the event of killing. The former lasts very little time, while the latter takes the whole interval between the trying and the dying. Silver (Reference Silver2018) recently proposed that the time of the killing is vague, and thus that there is something wrong in the presuppositions of the other authors I just mentioned.

I am not sure what to think about the time of a killing. Perhaps there is a right answer, or maybe the question is ill-formed, as Pietroski suggests.Footnote ⁵ Since the difficulties in locating both killings and answering machine utterances are similar, I suspect that we are facing a common issue about action in general. Once we understand the mechanism of the examples discussed, other examples can be easily constructed: suppose that I set a timer on my very fancy kettle, which will boil the water 16 hours after I set the timer. Suppose that all goes well, and the water does get boiled. We can ask a question like Thomson’s: when did I boil the water? Was it when I set the timer? Was it when the water was done getting to the boiling point? The whole period?

Furthermore, in neither case will the right answer be found merely by testing intuitions. If it were that simple, we would have found it already. Instead, I suspect that giving the right answer involves carefully weighing many different types of considerations, some about intuitions, some about the metaphysics of actions and events.

I am not implying that speech acts are in all respect like killings. One may think that speech acts take place exactly when something normative changes about the speaker, in that they take on certain obligations, e.g. following Austin ([1962] Reference Austin and Edited by1975, 116) and especially Sbisà (Reference Sbisà2013). But this answer will not allow for all the variation that we encounter in the ways that “now” is used. Consider this case from Cohen (Reference Cohen2013): A sends a postcard while on holiday in Venice, bearing the message “I wish you were here”. Intuitively, “here” picks out Venice, not the place where B reads the message. Another type of case yields different intuitions, which depend on the time and place of reading, not of writing. Someone could record a message, to be sent after one’s death, which says “Today you all received a phone call telling you to come to my lawyer’s office” (see Sherman (Reference Sherman2015)). “Today” seems to pick out the day of the reading, not the day of the recording. Did the postcard writer make their assertion in Venice, which the simple view would require? And did the will writer make their assertion posthumously in the lawyer’s office? If not, the simple view is wrong.

4. Detonation Conditions

I have argued that the metaphysics of speech acts is unlikely to give us the right inputs for a metasemantics of answering machine utterances that contain tokens of “now”. It would be tempting to take this as showing that all token reflexive views face this difficulty. Fortunately, we can appropriate some lessons from the vast literature on answering machines.

The first step is to think about what it is for a token to be used in a speech act, and why we should restrict our attention to tokens as used. First, a single token may be used to make multiple speech acts, and thus may acquire several referents. For example, a note saying “I’ll be back at noon” may be used by several people at different times, to say of each of them that they will be back at noon on the relevant day. So the token itself, independently of its uses, cannot be said to have any particular referent (see Kaplan (Reference Kaplan, Reichenbach, Kamlah and Reichenbach1999) and Perry (Reference Perry2001, 46).

Second, as O’Madagain (Reference O’Madagain2014) points out, one may create a token and not use it for its semantic functions, e.g. as part of a cursive writing exercise. Unused tokens do not get a referent.

So we need to talk about tokens as used, which is the first idea I will borrow. Using my own terminology, O’Madagain (Reference O’Madagain2014) proposed that we think of tokens as needing to be activated in order to function semantically. So the production of a token is separable from its starting to work semantically, and the starting requires deployment in a speech act.Footnote ⁶

But we still have our original problem: when exactly does “now” get activated and start to work semantically if used in an answering machine message, or in a letter sent by mail, and when does it refer? And how do we explain the fact that in some cases it picks out the time of the writing of the letter, but, in other cases, it picks out the time of the reading? Should we also say that the time of the speech act differs in similar ways? O’Madagain claims that the speaker intends to pick out a particular moment in time, and “now” gets activated exactly when the speaker wants it to be activated. And once active, it does what its character has it do: it picks out the current time, that is, the time of its activation.

In defending this idea, O’Madagain (Reference O’Madagain2014, 75) appeals to a broader view, which he calls “minimal intentionalism”: if $ \Phi $ is a semantic token meaning p, then S intends that $ \Phi $ expresses p. It is unclear how O’Madagain thinks of the structure of the relevant intention. One reading is that the speaker must have a singular intention about a particular time in order for a token to pick out that time. This view strikes me as incorrect, given plausible assumptions about singular intentions: the writer of the postcard may not know when the card will be read, but may still use “now” to pick out the time of its reading, whenever that may be.

A more plausible reading is that “now” may not go against the speaker’s intentions. This is better, because presumably the card writer did intend for his token of “now” to pick out the time of its reading, and it does just that. But we now run into what King (Reference King2014) called “the problem of conflicting intentions”: the speaker may have intended both that his token of “now” pick out the time of its reading, and that it pick out Jan 22, 2018, because that is when the card was intended to be read. Which intention is the semantically relevant one?

Fortunately, we do not need to settle the matter here. The view I propose below can be easily adapted by plugging in one’s preferred answer, e.g. Bach (Reference Bach1992), Bianchi (Reference Bianchi2014), Predelli (Reference Predelli1998), Predelli (Reference Predelli2005), Stojnic, Stone, and Lepore (Reference Stojnic, Stone and Lepore2013), Stokke (Reference Stokke2010).

Leaving aside intentionalism, there are two other problems with O’Madagain’s view. First, it allows tokens of “now” to pick out just about any moment of time, in any way the speaker may see fit, which seems too lax. Second, when I write a token of “now” in a postcard, it is already activated, even if this is a case where the token will only pick out the time of the reading, not the time of the writing. This token is already of a different kind than tokens made as part of calligraphy or diction exercises. It is a token that is already hard at semantic work; it is deployed, but isn’t referring yet, and will only refer if some conditions are met. We need more details.

The next step is to make use of the idea in Egan (Reference Egan2009) that there are several detonation conditions available for “now”, again rephrased with my terminology. With apologies for the extended militaristic metaphors, let me list a few.

In its most straightforward use, a token of “now” picks out the moment of its production. To use an example mentioned above, I may write on a postcard “the Sun is now finally out”, and thus pick out the time of my writing that very message. Like a bomb with a very short fuse, its detonation condition is so trivial that, if one focuses only on instantaneous communications, it may be almost invisible to the semanticist’s eye.

“Now” may also be used as a trigger bomb: its tokens may get detonated upon the right kind of contact. A token of “now” in a postcard may pick out the time of reading because of the triggering of the detonation condition of that particular token. There are different kinds of trigger bombs, depending on the features of the required trigger: some are only detonated when in contact with a particular person. Suppose, for instance, that the postcard contains the following sentence: “you won’t believe how good the wine is over here”. It seems to me that in this case “you” will get a referent only if the person reading the message is the intended person. Censors, or the nosy neighbor, may read the postcard, but they don’t get to be the addressee, and hence don’t get to be the referent of that token of “you”.

An interesting question is whether the (intended) addressee gets to be the referent of a token of “you” even before they read the note, or even if they don’t ever read the message. I think the latter is possible: I can write a letter to my great-great-grandparents, though they were dead before I was born, and I can use “you” to refer to one of them.

What about in our postcard case? My inclination is to think of this as a speech act that only takes place if the postcard is read. Before it is read, there is no speech act, and hence there is no activated token of “you” to be doing any referring.

Other uses may be triggered by whoever reads the tokens. One such interesting case discussed in Egan (Reference Egan2009, sec. 3) is “you”, as used on a billboard saying “Jesus loves you”.Footnote ⁷ Intuitively, there is nothing that the token picks out until it is read, and then it picks out whoever reads the message (although Zakkou (Reference Zakkou2017) disagrees). Egan likens this detonation condition to a shotgun: there is one broadcasting event (a deployment, in my terminology), but a new proposition gets expressed whenever the message is read. Because Egan is working with a Lewisian notion of a context, which is a location in space, time, and possible world space, his central claim is that some expressions are sensitive to both the speaker context and the addressee context. Since I think in terms of speech acts, I take it to be a case in which some features of the speech act are addressee-related, and they play a role in securing a referent for a used token of “you”.

An additional benefit of discussing shotgun cases is that we can see detonation conditions do more work than just determining a referent for token-reflexive expressions. Suppose that nobody reads the “Jesus loves you” sign. Then not only does that token of “you”, thus deployed, get no referent; nobody gets addressed either. The latter fact is independent of the words contained in the sign; if the sign had said “Bears are mammals”, and nobody had read it, it would still be the case that nobody would have been addressed by that message (recall that I am imagining a case where the message only works when read). In fact, I would go further, and claim that no speech act takes place if nobody reads the message.Footnote ⁸

It goes beyond the scope of this article to list all types of detonation conditions for all kinds of context sensitive expressions, a work that should be done one token-reflexive expression at a time. But we have enough to get an official formulation of the basic view, applied to “now”:

This is not quite the view I stated at the beginning of the article. The full view, as applied to “now”, specifies what determines the detonation conditions:

The main goal of this article is to present and defend the detonation view in general, as a solution to the (new) answering machine problem. In this article, I offer no arguments for the stronger intentionalist thesis, which specifies what determines the detonation conditions, since issues about intentionalism abound, and are common to any area where the speaker’s intentions are said to play a role. That said, I believe that the speaker’s intentions fully determine which conditions apply in a particular case. I also believe that for every expression, there is a list of detonation conditions, which is known by speakers of that language, and the speaker chooses one detonation condition from that list. Just like in any discussion of meaning, we must also note that innovation is, of course, possible, so the speaker may attempt to introduce a new detonation condition, for example in order to deal with new means of communication. But this should not obscure the fact that when the speaker attempts to use a novel detonation condition, they are attempting to add to the standing meaning of that expression.

The full view, then, counts as intentionalist, in the sense that speakers’ intentions wholly determine detonation conditions. But it is not a piecemeal intentionalism that has speakers pick a referent for each use as if from a blank slate. Rather, it is rule-type intentionalism, where speakers choose which of the detonation condition types gets to be applied when they use a token.

There are a number of rivals to the full view, each of which would count as implementations of the basic view. First, there is the piecemeal intentionalism just mentioned. Second, there is the other extreme, according to which the speaker’s intentions play no role in determining detonation conditions. And third, the widest category, are hybrid views, according to which intentions play some role, but only along with other, speaker-unrelated conditions. Each of these views has proponents; not about detonation conditions, of course, since this is a novel notion, but rather about similar debates about the role of intentions in the referential mechanism of demonstratives.

In the next section, I will defend both the basic and the full view against some objections. This will also allow me to compare these views to solutions to the more classical answering machine problem, as it relates to Kaplan (Reference Kaplan, Almog, Perry and Wettstein1989b).

5. Other Views

The closest to my proposal is the shifty character view of Michaelson (Reference Michaelson2014). Following Kaplan (Reference Kaplan, Almog, Perry and Wettstein1989b), Michaelson takes the character of an expression to be the rule that yields a referent relative to a context. Michaelson’s innovation is to claim that answering machine cases show that indexicals have not just a fixed character, but rather that their character shifts as a function of the medium of communication. For example, “now”, when spoken, refers to the time of speech. But when written in a note, it refers to the time of reading, not of writing. Michaelson’s view is thus an attempt to fix the answering machine problem as it arises for Kaplan’s classical view, while using Kaplan’s conceptual framework.

In my view, Michaelson’s proposal has two main problems.

First, my intentionalist stance is counter to Michaelson’s non-intentionalist one. This is a larger issue than I can address here, but let us look at two cases that highlight the advantages of an intentionalist stance. The first case, from Bianchi (Reference Bianchi2014, 488), is about a related matter: the illocutionary force of speech acts. Homer writes a note, saying “Don’t leave”, and leaves it for the butler on the butler’s bed. Marge finds it and thinks it was meant for her, begging her not to leave. I agree with Bianchi’s intuitions that this was a case in which the butler, not Marge, was being addressed. In fact, I would go a bit further, and point out that the answering machine-ness of the case is not essential here. Suppose that Homer, Marge, and the butler are in the same room, and Homer says “Don’t leave”. Marge may well take Homer to be addressing her, and she would be wrong if Homer was addressing the butler, a matter that I take to depend on Homer’s intentions.

My intuitions about the referents of indexicals are just as intentionalist. Suppose the note said “I noticed your pen was leaking, so I bought you a new one”. My intuition is that the referent of the second person pronouns is the addressee of the note, namely the butler, even if Marge reads the note before the butler. And, again, it is not important that the message be in note form. Just as in the previous case, they could all be talking in the same room, and nothing essential would change.

Second, there is no reason to think that each channel comes with exactly one meaning rule for each indexical. Identifying the medium is not irrelevant to figuring out the speaker’s intentions, and some media may well be characteristically associated with just one detonation condition. For instance, in face to face communication, tokens of “now” typically pick out the time of their deployment, because there is no other option available. But, as Michaelson (Reference Michaelson2014, 527) acknowledges, “today” written in a will may be used to pick out the day of its writing, or the day of its reading-as-intended (e.g. after the writer’s death, by the lawyer, in front of the family, etc.). Michaelson’s suggestion is that in this case no single convention has emerged for wills. I agree with this claim, but I see no pressure on users to choose just one condition per medium.

This discussion of Michaelson’s view prompts the following question: what is the relation between detonation conditions and Kaplan’s notion of character as reference-determining meaning? I confess that I am not sure. Kaplan’s notion of character has the same job as detonation conditions, since they both prescribe the way expressions get their referents. I also agree with Kaplan’s view that “I” picks out the speaker, “now” the current time, etc. But here lies one issue: I seem to claim that there is just one meaning of “now”, but several detonation conditions. So which one is the meaning, or character, of “now”?

I suggest that this is the wrong question to ask (as do the various metarule views, such as Corazza, Fish, and Gorvett (Reference Corazza, Fish and Gorvett2002) and Michaelson (Reference Michaelson2014)). Speakers, when learning the meaning of “now”, need only learn that it picks out the current time. But that cannot be quite right, because there are no universally obvious intuitions about the current time of a speech act. So it picks out the (current) time of its detonation. Given my claim that detonation conditions are selected from a relatively small list, the speaker must also learn these conditions. Which is the meaning of “now”? No single answer suffices. To tell the story of how one token of “now” picks out a moment in time requires talking about how it picks out the current time, what kind of detonation conditions it had, and whether those conditions were met. All this deserves to be seen as part of the meaning of “now”. The meaning of “now” will thus have much in common with the meaning of the French “maintenant”, because they both pick out the time of the detonation. So learning a new language need not always involve learning new detonation conditions.

This allows me to answer an objection from O’Madagain (Reference O’Madagain2021): multiple-rule theories are only plausible if there is a small number of rules for each expression. If every new case is to be explained by positing a new rule, language learning and language comprehension would be unexplained. In reply, there is a sense in which my theory is a single rule theory: “now” picks out the time of detonation. This does not exhaust the meaning / character of “now”, nor does it push back the answer impermissibly because the list of detonation conditions is short, and it contains the natural options already discussed for killing: “now” can pick out the time of deployment, or the time of receipt. Admittedly, “receipt” amounts to various things, depending on whether it is a trigger use or a shotgun use, but that list is not long either. As we see below, the speaker may also have the option of picking out the intended time of receipt, even if receipt does not actually take place. Whether that makes it on the list or not, the list is contained, and learnable.

Another close relative of my view is the deferred utterance view of Sidelle (Reference Sidelle1991), recently spelled out and defended in Briciu (Reference Briciu2018). The central idea is that in answering machine cases the speaker arranges for a deferred utterance to happen at some later time or place. Inasmuch as “utterance” is meant to pick out a locutionary or illocutionary act, I take it that this view is shown to be wrong by the lack of intuitions about the time or place of answering machine speech acts. But I do agree that a speech act takes place, and that the speaker is setting up the conditions for it to happen. So the only disagreement is in the specification of the detonation conditions: I take these to be separate from the time and place of the speech act, if there even are such things as the time and the place of the speech act.

The disagreement is clearly seen in a case discussed in Briciu (Reference Briciu2018, 46): a station master records a message, saying “The train now approaching does not stop here”, to be played at a later time. Suppose that the message is played just when a train is coming into the station, and the train does stop. Briciu and I agree that the station master can be accused of having asserted something false. Briciu takes this to show that the speech act happens exactly during the playback. I disagree: it only shows that the detonation conditions were met, and that the station master is responsible for the message. But I have no such clarity on the time of the speech act, so I would separate the question about detonation conditions from the question about the timing of the act.

The final issue I would like to address is Predelli’s “threatening note”, as Perry (Reference Perry2003) calls it. Here is the case again: E leaves a message for F, saying “I am not here now”. E believes that F will get home and read the message at 6pm, and E intends for “now” to pick out 6pm. F only reads it at 10pm.Footnote ⁹

This is a hard case, because intuitions seem to vary with respect to the referent of “now”. Predelli claims that it refers to 6pm (and thus it would be like a timed bomb, with no other detonation conditions). Briciu (Reference Briciu2018, sec. 3.2) claims that it picks out nothing. Dodd and Sweeney (Reference Dodd and Sweeney2009, 341) say that it picks out 10pm.

There are also hard questions about the purported speech act. Sidelle (Reference Sidelle1991) and O’Madagain (Reference O’Madagain2014) claim that there is an utterance at 6pm, although no event takes place then. Briciu (Reference Briciu2018) claims that E’s intention misfires, so no utterance ever takes place. Instead, our intuition that “now” picks out 6pm is explained by our intuitions about a merely possible utterance, which would have taken place had the note been read at the right time.

I have no clear intuitions about the referent of “now” here. I also have no clear intuition whether, much less when, such a speech act takes place.

In addition, the case is underspecified. Here are two different ways that the case might go. First: E deploys a token of “now”, with the intention that it detonate at 6pm, when it is read by F. I assume that E’s intention is structured exactly as I described it. In this case, I have no clear intuition about the time of the speech act, or the referent of that token of “now”.

Second: E intended that the token of “now” detonate upon being read by F, and further believed that it would be read at 6pm, where that belief is not part of the intention. Under this specification, I do have the clear intuition that we get 10pm as a referent.

Speaking of detonation conditions allows me to explain my lack of intuitions under the first specification: there just isn’t an established practice of picking out such detonation conditions for “now”. We certainly may intend our messages to be read by particular people, or at particular times. But we have no practice of doing what E is trying to do. So I lack the relevant intuitions. One option is to think of this as a faulty detonator case. However, Predelli does have the 6pm intuition. My theory can explain his intuition too: perhaps he takes it that such a practice exists, or that it can be set up by E. I am simply not sure, because it seems to me too hard for the speaker to be able to figure it out reliably enough, so it wouldn’t be a very useful means of communication.Footnote ¹⁰ If the practice gets established, maybe by some further technological invention that makes it natural, I would acquire the relevant intuitions. Until then, I remain on the fence, because it just doesn’t seem like a reasonable practice to me, since it makes communication quite hard to accomplish. But at least the fence is one that I can describe with the conceptual apparatus developed in this article, which serves to make the sitting more comfortable.